NEWS AND V IEWS

ferences in average copy number between the HapMap populations, notably including CCL3L1, which is involved in HIV suscep-tibility9. Unfortunately, there may be diffi-culties in detecting selection on CNV loci: the standard pattern of a selective sweep in surrounding SNPs may not be very strong, as higher rates of mutation make selected CNV alleles more likely appear on multiple haplotypes10. To adjust for this, scans for selection between populations should there-fore include direct targeting of CNV loci, as they are excellent candidates for selection

but may not be well tagged by surrounding variation.

The example of amylase might also be representative of the importance of diet as a widespread selective pressure in human evo-lution. For example, it is already well known that the selective sweep associated with lac-tase persistence is one of the strongest signals of selection in the genome11,12, and multiple populations have evolved lactase persistence independently4. So, as far as our evolutionary history is concerned, it seems we really are what we eat.

1. Perry, G et al. Nat. Genet. 39, 1256–1260 (2007).2. Laden, G. & Wrangham, R. J. Hum. Evol. 49, 482–498

(2005).3. Thompson, E.E. et al. Am. J. Hum. Genet. 75, 1059–1069

(2004).4. Tishkoff, S.A. et al. Nat. Genet. 39, 31–40 (2007).5. Helgason, A. et al. Nat. Genet. 39, 218–225

(2007).6. McCarroll, S.A. et al. Nat. Genet. 38, 86–92 (2006).7. Stranger, B.E. et al. Science 315, 848–853 (2007).8. Redon, R. et al. Nature 444, 444–454 (2006).9. Gonzalez, E. et al. Science 307, 1434–1440

(2005).10. Pennings, P.S. & Hermisson, Mol. Biol. Evol. 23,

1076–1084 (2006).11. Bersaglieri, T. et al. Am. J. Hum. Genet. 74, 1111–1120

(2004).12. Voight, B.F. et al. PLoS Biol. 4, e72 (2006).

Rhythm is not enoughBernhard Horsthemke

MAGEL2 is located in a cluster of imprinted genes on human chromosome 15 that is implicated in Prader-Willi syndrome (PWS). A new study shows that mice deficient for this gene show altered behavioral rhythmicity that resembles some features of PWS.

The jet lag we feel after crossing several time zones reminds us that many of our bodily functions are under the control of a biologi-cal clock. The central component of this clock, located in the hypothalamus, is a transcrip-tional-translational feedback loop that gen-erates circadian (∼24 h) oscillations (Fig. 1). The clock can be reset by light via the input pathway, and the output pathway translates the oscillations into physiological and behavioral rhythms. As shown by Kozlov et al. on page 1266 of this issue1, the imprinted gene Magel2 seems to be an important component of the output pathway in mice.

The hypothalamus and PWSIn a screen for cycling transcripts in the hypo-thalamus of mice, Panda et al.2 found that a relatively small number of output genes are directly regulated by the oscillator, and Magel2 is among them. The human ortholog, MAGEL2, is located on the proximal long arm of chromo-some 15 (15q11-q13) and is expressed from the paternal allele only. Lack of a paternal copy of 15q11-q13 causes PWS, a neurobehavioral disease characterized by neonatal muscular hypotonia and failure to thrive, hyperphagia and obesity starting in early childhood, hypo-gonadism, short stature, behavioral problems and mental retardation. As MAGEL2 is highly

expressed in the hypothalamus, and a hypo-thalamic defect most likely underlies PWS, MAGEL2 deficiency may account for some of the clinical features of PWS.

Kozlov et al.1 generated mice that do not express Magel2. These mice were weaned at a somewhat lower frequency than expected, but they did not show any substantial altera-tions in size or weight, even by two years of age. However, although the mice had a normal circadian rhythm, they ran noticeably less dur-ing the night (when mice should be awake) and had increased activity during the day (when mice should be asleep). These changes were associated with reductions in food intake and male fertility, and they establish Magel2 as a circadian-output gene.

Circadian rhythmicity and imprintingThe work raises three important questions: (i) how does Magel2 translate the oscillations into behavioral and physiological rhythms,

(ii) why is a circadian-output gene subject to genomic imprinting and (iii) what is the role of MAGEL2 deficiency in PWS?

The function of the Magel2 protein is unknown, but Kozlov et al.1 found reduced levels of the neuropeptides orexin A and B in the hypothalamus of Magel2-deficient mice, whereas the levels of prepro-orexin, from which the orexins are derived, were elevated. Orexin-expressing neurons project throughout the brain, particularly in regions implicated in regulating sleep-wakefulness, body temperature and feeding. The authors speculate that Magel2 may modulate orexin signaling by affecting the post-translational processing of prepro-orexin to orexin. This effect, however, must be indirect, because Magel2 is not a prepropeptide-convert-ing enzyme.

Magel2 maps to a chromosomal domain that is differentially marked (imprinted) by DNA methylation during oogenesis and spermato-genesis. Consequently, the maternal allele is

Bernhard Horsthemke is at the Institut für Humangenetik, Universitätsklinikum Essen, Germany. e-mail: bernhard.horsthemke@uni-due.de

Input pathway Output pathway(Magel2 and

other proteins)

Circadianoscillator

Figure 1 An autonomous, self-sustained oscillator acts as the core of the biological clock. The clock can be reset by light and controls various bodily functions, including wakefulness and sleep.

1190 VOLUME 39 | NUMBER 10 | OCTOBER 2007 | NATURE GENETICS

Kim

Cae

sar

©20

07 N

atur

e P

ublis

hing

Gro

up

http

://w

ww

.nat

ure.

com

/nat

ureg

enet

ics

NEWS AND V IEWS

silenced. Genomic imprinting coevolved with the placenta and seems to regulate the alloca-tion of resources between mother and fetus. As proposed by the genetic conflict theory3, the paternal genome favors growth of the fetus at the expense of the mother, whereas the maternal genome restricts fetal growth to preserve resources for future pregnancies. As the hypothalamus regulates feeding and energy metabolism, it does not come as a sur-prise that several imprinted genes, including Magel2, are expressed in hypothalamic cells. However, it is a puzzle why a circadian-output gene should be imprinted. A priori, one would have thought that circadian rhythmicity would be more important after birth than before birth and would therefore be a relatively unimport-ant issue in the parental conflict over maternal resources during fetal growth.

Perhaps Magel2 is just an innocent bystander. By comparing the orthologous genes in humans, marsupials, platypus and other animals, Rapkins et al.4 found that human 15q11-q13 was assembled from non-imprinted chromosomal components and that imprinting of this region was acquired relatively recently in the placental lineage. Magel2 and its human ortholog MAGEL2 are intronless genes derived from an ancestral gene on the X chromosome,

probably by retrotransposition. The retrotrans-posed copies then came under the control of a domain-wide imprinting center in the region. Thus, there may be no specific selective advan-tage for imprinting of Magel2 and MAGEL2.

Although MAGEL2 is a good candidate gene for some of the features of PWS, for example, daytime sleepiness and nocturnal sleeping problems, the mice generated by Kozlov et al.1 did not develop hyperphagia and obesity, which are hallmarks of PWS. At first glance, this suggests that MAGEL2 is not the principal PWS-associated gene. However, even mice with a large deletion resembling that found in the majority of individuals with PWS5 and mice with an imprinting defect across the whole domain6 do not develop hyperphagia and obesity. First, this reminds us that mice and humans are different. However, more sophis-ticated studies of the mice are necessary. For example, it will be important to determine the ratio of lean versus fat mass in these animals, because the mice may be relatively fat without being obese. Additionally, there are as yet no human data suggesting that MAGEL2 has an important role in PWS. In fact, studies of rare individuals with PWS who carry a balanced translocation7,8 that disrupts 15q11-q13 point to the presence of a candidate gene a consider-

able distance away from MAGEL2. This criti-cal region harbors several small RNA genes7,9, and the research community eagerly awaits the description of mice and humans carrying deletions of just these genes. However, even if the small RNA genes are found to be the prin-cipal genes involved in PWS, it is possible that they interact with other genes in the region, including MAGEL2, and that the full PWS phenotype results from the dysregulation of several genes.

In summary, the mice generated by Kozlov et al.1 provide a possible link between circa-dian function in the hypothalamus and the Prader-Willi phenotype and emphasize the importance of the output pathway of our bio-logical clock. Rhythm alone does not make a song.

1. Kozlov, S.V. et al. Nat. Genet. 39, 1266–1272 (2007).

2. Panda, S. et al. Cell 109, 307–320 (2002).3. Moore, T. & Haig, D. Trends Genet. 7, 45–49 (1991).4. Rapkins, R.W. et al. PLoS Genet. 2, e182 (2006).5. Gabriel, J.M. et al. Proc. Natl. Acad. Sci. USA 96,

9258–9263 (1999).6. Yang, T. et al. Nat. Genet. 19, 25–31 (1998).7. Runte, M. et al. Hum. Mol. Genet. 10, 2687–2700

(2001).8. Gallagher, R.C., Pils, B., Albalwi, M. & Francke, U. Am.

J. Hum. Genet. 71, 669–678 (2002).9. Cavaillé, J. et al. Proc. Natl. Acad. Sci. USA 97,

14311–14316 (2000).

How microRNAs choose their targetsIvo L Hofacker

The growing list of known microRNAs is only as useful as our ability to identify the mRNA targets they control. A new study stresses the role of messenger RNA structure in microRNA target recognition and suggests that binding of the RNA-induced silencing complex is largely controlled by the thermodynamics of RNA-RNA interactions.

Ivo L. Hofacker is at the Institute for Theoretical Chemistry, University of Vienna, Währingerstraße 17, A-1090 Wien, Austria. e-mail: ivo@tbi.univie.ac.at

MicroRNAs (miRNAs) provide a unique regu-latory layer that is essential for the regulation of gene expression in most metazoans. These small RNAs of about 22 nucleotides in length guide the RNA-induced silencing complex (RISC) to target sites, usually located in the 3′ UTR of select mRNAs. Binding of RISC leads to suppression of translation, or even degrada-tion, of target mRNAs. The beauty of the sys-tem lies in the use of short RNAs, which can be produced cheaply and can evolve quickly, to regulate a large and complex protein machin-ery. Over the past few years, the number of

known miRNAs collected in the miRBase1 reg-istry has grown rapidly to over 5,000, including 533 human miRNAs, and recent studies suggest that the repertoire of human miRNAs is not yet remotely exhausted2.

Experimental identification of target mRNAs is difficult, however, and Tarbase3 currently lists only 130 verified targets for 56 human miRNAs. Consequently, computational methods for miRNA target prediction are in high demand. The task is complicated by the fact that complementarity between miRNAs and target mRNAs is generally far from per-fect: often, only a short ‘seed’ region of 6–8 bases at the 5′ side of the miRNA is perfectly base-paired to the mRNA. Consequently, a simple search for sequence complementarity is expected to produce many false-positive hits. Moreover, most miRNAs seem to have several

(possibly hundreds) of targets, some of which will be targeted more strongly than others. Ideally, one should therefore predict not just targets, but also the expected degree of trans-lational suppression.

To overcome this problem, methods cur-rently popular, such as PicTar4 and TargetScan5, rely heavily on evolutionary conservation of the target site. Although conservation is a powerful way to improve the detection signal, it is clearly not useful for the (potentially large) class of species-specific miRNAs that have recently been found6. It is also unsatisfying, as it leaves open the mechanism by which miRNAs find their target mRNAs in the cell. On page 1278 of this issue, Kertesz et al.7 add a new dimension to miRNA target prediction that improves per-formance and suggests a new role of RNA-RNA interactions in regulating RISC binding.

NATURE GENETICS | VOLUME 39 | NUMBER 10 | OCTOBER 2007 1191

©20

07 N

atur

e P

ublis

hing

Gro

up

http

://w

ww

.nat

ure.

com

/nat

ureg

enet

ics